Table 4.

Ranking-based evaluation metrics of 5 embedding-based link prediction models on the Alzheimer’s Knowledge Base knowledge graph. Metrics are derived from the likelihood of existing (known) links being predicted by the models. Higher scores indicate better performance.

Model name	Hits@1	Hits@3	Hits@10	MRRa
RotatE	0.126 b	0.220	0.358	0.202
TransE	0.046	0.097	0.198	0.097
DistMult	0.027	0.056	0.126	0.061
ComplEx	0.074	0.142	0.263	0.136
ConvE	0.002	0.005	0.013	0.006

^aMRR: mean reciprocal rank.

^bItalicized values indicate maximum scores within a given column.